Some of construction machines (e.g., large hydraulic shovels) operating in mines, etc. are required to operate 24 hours a day almost without stopping and 365 days a year consecutively. Such a construction machine has to be kept in perfect condition by previously doing maintenance such that the machine does not stop due to an abnormality. In general, a specialized maintenance personnel periodically conducts inspection to check whether or not there exists an abnormal part. When an abnormality is found, necessary maintenance work is conducted to the machine, by which satisfactory machine conditions are maintained. However, it is necessary to stop each machine in order to carry out the check and maintenance. Thus, for an operator who wants to make machines operate continuously, the check and maintenance work would be cumbersome in operation if the machines are kept in fine condition.
Therefore, abnormality diagnosis technologies have developed for measuring conditions of a machine using various sensors and checking whether or not the machine has an abnormality. Preventive maintenance, which prevents machine failure by previously detecting an abnormality by a diagnostic technology before the machine fails or stops and conducting maintenance as countermeasures in early stages, is being regarded as more and more important.
Meanwhile, even though machine manufacturers are energetically developing diagnostic algorithms for the abnormality diagnosis, there are cases where appropriate judgment is impossible as a result of difficulty in the algorithm development. The appropriate judgment is difficult since the experimental environment in which experiments were conducted for the algorithm development differs from the operating environment and operational configurations in which the user actually uses the machine.
In consideration of the above problem, inventions aiming to make a judgment using measurement results in the actual environment have been made. For example, in a technique described in Patent Literature 1, average temperature is calculated as the average of outputs of temperature sensors placed at the cylinders of the engine. When the difference between the temperature of a cylinder and the average temperature reaches a preset value, the cylinder is judged to have failed.